LISTSERV at the University of Georgia
Menubar Imagemap
Home Browse Manage Request Manuals Register
Previous (more recent) messageNext (less recent) messagePrevious (more recent) in topicNext (less recent) in topicPrevious (more recent) by same authorNext (less recent) by same authorPrevious page (December 2004, week 4)Back to main SAS-L pageJoin or leave SAS-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:         Wed, 22 Dec 2004 13:16:10 -0800
Reply-To:     cassell.david@EPAMAIL.EPA.GOV
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         "David L. Cassell" <cassell.david@EPAMAIL.EPA.GOV>
Subject:      Re: Negative binomial regression with weighted data
In-Reply-To:  <200412220214.iBM2Elwf014103@listserv.cc.uga.edu>
Content-type: text/plain; charset=US-ASCII

Alain Girard <alain.girard@UMONTREAL.CA> wrote: > I want to do a negative binomial regression and i have weighted data. > > In genmod the weight statement is a scale weight. It's look not realy a > subject weight. > > I think to use the frequency statement in genmod. But all my weight are > between 0 and 1, the frequency statement remove the subject with freq. > lower than 1 (and i lost all my subject !!!).

First of all, the FREQ statement will round down. If you give it a frequency of 100.4, it will truncate to 100. If you give it a frequency of 0.4, it will truncate to zero. A zero frequency (or a zero weight) will ensure that the given record won't be used in the analysis.

Next, I don't see how your numbers can be weights at all. If your values are between 0 and 1 (not including 1, since none of your subjects were retained when you used the FREQ statement), then they have no meaning as weights. They certainly couldn't be sampling weights.

Perhaps you have some values which represent some manner of distribution of responses. If so, you still cannot use them, since there is no way to reclaim any sense of the sample size from a 'distribution' of data.

Perhaps they have been 'scaled' first (a bad idea). Perhaps you know a set sample size, and these 'weights' are the real weights, divided by your n. If so, multiply back by n! Get the real weights (or frequencies, or whatever they are) and go from there.

Please write back (to the list, not to me personally) and explain what your 'weights' really are, and what you hope to achieve with them.

HTH, David -- David Cassell, CSC Cassell.David@epa.gov Senior computing specialist mathematical statistician


Back to: Top of message | Previous page | Main SAS-L page